PentAGI is an open-source AI agent system for autonomous penetration testing and security research in authorized environments. It combines language-model reasoning with security tools and isolated execution so an agent can plan, investigate, run commands, collect evidence, and iterate through multi-step assessment tasks.

The platform integrates a range of offensive-security and reconnaissance capabilities, browser or search functions, terminal tools, memory, and multi-agent-style task handling. Its interface and orchestration layer are designed to keep long security workflows organized while preserving records of what the agent attempted and discovered.

PentAGI is intended for security professionals, labs, and controlled testing where users have permission to assess the target systems. It should be treated as specialized security tooling rather than a general assistant, with deployment and use governed by the same authorization and operational controls applied to conventional penetration-testing tools.